Position: Retire the "Positive Backdoor" Label—Secret Alignment Requires Strict and Systematic Evaluation

Jianwei Li, Jung-Eun Kim
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:171213-171239, 2026.

Abstract

This position paper argues that the AI/ML community should stop overclaiming and retire the label “positive backdoor”, and instead treat trigger-activated hidden behaviors as Secret Alignment. Crucially, protective claims based on Secret Alignment should be presumed not secure by default unless supported by rigorous, standardized evaluation. The Private AI era, enabled by open-weight LLMs and accessible training/inference stacks, turns language models into privately owned digital assets, creating security concerns around unauthorized access, model theft, and behavioral misuse. Recently, a line of work framed as “positive backdoors” has been proposed to address these challenges. To ground our position in evidence, we unify these proposals as covert trigger-behavior associations for access gating, ownership attribution, and safety enforcement, and evaluate three representative applications across six core properties: effectiveness, harmlessness, persistence, efficiency, robustness, and reliability. Our results reveal substantial brittleness–especially in the confidentiality, integrity, and availability (CIA)–of trigger-behavior mappings often underrepresented by existing claims. We further relate these outcomes to behavior density and decision complexity, offering a behavioral lens for understanding deployment-time risks and motivating community-wide evaluation that makes Secret Alignment claims provable.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-li26kc, title = {Position: Retire the "{P}ositive Backdoor" Label—Secret Alignment Requires Strict and Systematic Evaluation}, author = {Li, Jianwei and Kim, Jung-Eun}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {171213--171239}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/li26kc/li26kc.pdf}, url = {https://proceedings.mlr.press/v306/li26kc.html}, abstract = {This position paper argues that the AI/ML community should stop overclaiming and retire the label “positive backdoor”, and instead treat trigger-activated hidden behaviors as Secret Alignment. Crucially, protective claims based on Secret Alignment should be presumed not secure by default unless supported by rigorous, standardized evaluation. The Private AI era, enabled by open-weight LLMs and accessible training/inference stacks, turns language models into privately owned digital assets, creating security concerns around unauthorized access, model theft, and behavioral misuse. Recently, a line of work framed as “positive backdoors” has been proposed to address these challenges. To ground our position in evidence, we unify these proposals as covert trigger-behavior associations for access gating, ownership attribution, and safety enforcement, and evaluate three representative applications across six core properties: effectiveness, harmlessness, persistence, efficiency, robustness, and reliability. Our results reveal substantial brittleness–especially in the confidentiality, integrity, and availability (CIA)–of trigger-behavior mappings often underrepresented by existing claims. We further relate these outcomes to behavior density and decision complexity, offering a behavioral lens for understanding deployment-time risks and motivating community-wide evaluation that makes Secret Alignment claims provable.} }
Endnote
%0 Conference Paper %T Position: Retire the "Positive Backdoor" Label—Secret Alignment Requires Strict and Systematic Evaluation %A Jianwei Li %A Jung-Eun Kim %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-li26kc %I PMLR %P 171213--171239 %U https://proceedings.mlr.press/v306/li26kc.html %V 306 %X This position paper argues that the AI/ML community should stop overclaiming and retire the label “positive backdoor”, and instead treat trigger-activated hidden behaviors as Secret Alignment. Crucially, protective claims based on Secret Alignment should be presumed not secure by default unless supported by rigorous, standardized evaluation. The Private AI era, enabled by open-weight LLMs and accessible training/inference stacks, turns language models into privately owned digital assets, creating security concerns around unauthorized access, model theft, and behavioral misuse. Recently, a line of work framed as “positive backdoors” has been proposed to address these challenges. To ground our position in evidence, we unify these proposals as covert trigger-behavior associations for access gating, ownership attribution, and safety enforcement, and evaluate three representative applications across six core properties: effectiveness, harmlessness, persistence, efficiency, robustness, and reliability. Our results reveal substantial brittleness–especially in the confidentiality, integrity, and availability (CIA)–of trigger-behavior mappings often underrepresented by existing claims. We further relate these outcomes to behavior density and decision complexity, offering a behavioral lens for understanding deployment-time risks and motivating community-wide evaluation that makes Secret Alignment claims provable.
APA
Li, J. & Kim, J.. (2026). Position: Retire the "Positive Backdoor" Label—Secret Alignment Requires Strict and Systematic Evaluation. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:171213-171239 Available from https://proceedings.mlr.press/v306/li26kc.html.

Related Material